Bacteria–phage infection network structure and genomic defence system content predict efficacy of a phage therapy cocktail against <i>Pseudomonas aeruginosa</i> from chronic lung infections

<i>Pseudomonas aeruginosa</i> chronic lung infections pose serious challenges for phage therapy due to high between-patient strain diversity and rapid within-patient phenotypic and genetic diversification, necessitating simple predictors of efficacy to streamline phage cocktail design. We quantified bacteria-phage infection networks (BPINs) for six phages against 900 <i>P. aeruginosa</i> clones previously isolated from 10 bronchiectasis infections (<i>n</i> = 90 isolates per patient). BPIN structure varied extensively between patients. The efficacy of the six-phage cocktail against these diverse <i>P. aeruginosa</i> populations was influenced by several factors. Cocktail efficacy increased with decreasing number and strength of individual resistances, as well as with increasing co-resistance modularity and phage dose. These results highlight simple BPIN metrics that could help guide the design of effective phage therapeutics. Resistance against some but not all the phages increased with higher number defence systems per genome, resulting in lower efficacy of the six-phage cocktail, suggesting that <i>P. aeruginosa</i> strains with fewer defence systems are better candidates for phage therapy. Overall, our findings suggest that 'off the peg' phage therapeutics are unlikely to be broadly effective against <i>P. aeruginosa</i> chronic respiratory infections, but that the design of personalised phage cocktails could be guided using simple BPIN metrics, and that defence systems per genome provide a useful rule of thumb for identifying highly treatable infections.This article is part of the discussion meeting issue 'The ecology and evolution of bacterial immune systems'.